Académique Documents
Professionnel Documents
Culture Documents
13
00:01:20,920 --> 00:01:23,580
goes through the microphone
transducer as a function of
14
00:01:23,580 --> 00:01:27,150
time, as representing
the speech signal.
15
00:01:27,150 --> 00:01:31,120
And we might see a typical
speech signal looking
16
00:01:31,120 --> 00:01:33,990
something like I've
indicated here.
17
00:01:33,990 --> 00:01:38,310
It's a function of time, in
this particular case.
18
00:01:38,310 --> 00:01:42,970
And the independent variable,
being time, is, in fact,
19
00:01:42,970 --> 00:01:44,290
continuous.
20
00:01:44,290 --> 00:01:47,780
And so a signal like this, we
will typically be referring to
21
00:01:47,780 --> 00:01:51,570
as a continuous time signal.
22
00:01:51,570 --> 00:01:55,180
Now, it also, for this
particular example, is a
23
00:01:55,180 --> 00:01:58,470
function of one independent
variable.
24
00:01:58,470 --> 00:02:06,660
And that will be referred to
as a one-dimensional signal
25
00:02:06,660 --> 00:02:10,470
corresponding to the fact that
there's only one independent
26
00:02:10,470 --> 00:02:13,890
variable instead of several
independent variables.
27
00:02:13,890 --> 00:02:18,696
So the speech signal is an
example of a continuous time,
28
00:02:18,696 --> 00:02:19,946
one-dimensional signal.
29
00:02:19,946 --> 00:02:21,520
30
00:02:21,520 --> 00:02:25,990
Now, signals can, of course,
be multi-dimensional.
31
00:02:25,990 --> 00:02:28,740
And they may not have, as their
independent variables,
32
00:02:28,740 --> 00:02:30,130
time variables.
33
00:02:30,130 --> 00:02:34,000
One very common example
are the examples
34
00:02:34,000 --> 00:02:36,080
represented by images.
35
00:02:36,080 --> 00:02:41,340
Images, as signals, we might
think of as representing
36
00:02:41,340 --> 00:02:46,020
brightness, as it varies
in a horizontal
37
61
00:04:23,600 --> 00:04:27,430
road system is somewhat obscured
by cloud cover.
62
00:04:27,430 --> 00:04:31,990
And what I'll want to show later
as an example of what a
63
00:04:31,990 --> 00:04:35,340
system might do to such a
signal, in terms of processing
64
00:04:35,340 --> 00:04:41,540
it, is an attempt to at least
compensate somewhat for the
65
00:04:41,540 --> 00:04:44,900
cloud cover that's represented
in the photograph.
66
00:04:44,900 --> 00:04:48,140
Although in terms of the
detailed analysis that we go
67
00:04:48,140 --> 00:04:52,550
through during the course, our
focus of attention is pretty
68
00:04:52,550 --> 00:04:54,810
much restricted to
one-dimensional signals.
69
00:04:54,810 --> 00:04:58,940
In fact, we will be using
two-dimensional signals, more
70
00:04:58,940 --> 00:05:03,390
specifically images, very often
to illustrate a variety
71
00:05:03,390 --> 00:05:06,090
of concepts.
72
00:05:06,090 --> 00:05:10,400
Now, speech and images are
96
00:06:44,800 --> 00:06:47,660
Incidentally, this particular
period was
97
00:06:47,660 --> 00:06:49,925
not chosen at random.
98
00:06:49,925 --> 00:06:54,260
It In fact captures a very
interesting aspect of stock
99
00:06:54,260 --> 00:07:00,640
market history, namely the stock
market crash in 1929,
100
00:07:00,640 --> 00:07:05,180
which, in fact, is represented
by the behavior of this
101
00:07:05,180 --> 00:07:08,720
discrete-time signal,
or sequence, in
102
00:07:08,720 --> 00:07:09,810
this particular area.
103
00:07:09,810 --> 00:07:12,960
So this dramatic dip, in
fact, is the stock
104
00:07:12,960 --> 00:07:16,610
market crash of 1929.
105
00:07:16,610 --> 00:07:21,340
Well, the Dow Jones weekly
average is an example of a
106
00:07:21,340 --> 00:07:24,520
one-dimensional discrete-time
signal.
107
00:07:24,520 --> 00:07:29,150
And just as with continuous
time, we had not just
108
00:07:29,150 --> 00:07:31,810
one-dimensional but
multi-dimensional signals,
109
00:07:31,810 --> 00:07:35,950
likewise we have
multi-dimensional signals in
110
00:07:35,950 --> 00:07:41,870
the discrete-time case where,
in that case, then, the
111
00:07:41,870 --> 00:07:43,900
discrete-time signal that
we're talking about, or
112
00:07:43,900 --> 00:07:48,160
sequence, is a function of
two integer variables.
113
00:07:48,160 --> 00:07:52,230
And as one example, this might,
let's say, represent a
114
00:07:52,230 --> 00:07:58,330
spatial antenna array where this
is array number in, let's
115
00:07:58,330 --> 00:08:01,730
say, the horizontal direction,
and this is array number in
116
00:08:01,730 --> 00:08:03,550
the vertical direction.
117
00:08:03,550 --> 00:08:06,070
Both classes of signals,
continuous-time and
118
00:08:06,070 --> 00:08:10,100
discrete-time, as I've
indicated, are very important.
119
00:08:10,100 --> 00:08:14,300
And it should be emphasized
143
00:09:49,770 --> 00:09:55,060
will do is attempt to exploit
some very specific, and as
144
00:09:55,060 --> 00:09:59,410
we'll see, very useful
system properties.
145
00:09:59,410 --> 00:10:04,160
To indicate what I mean and how
things might be split up,
146
00:10:04,160 --> 00:10:07,660
we could talk about systems, and
will talk about systems,
147
00:10:07,660 --> 00:10:10,300
that are linear.
148
00:10:10,300 --> 00:10:14,090
And we could divide systems,
basically, into systems that
149
00:10:14,090 --> 00:10:20,510
are either linear or nonlinear,
and we will, and
150
00:10:20,510 --> 00:10:24,590
also divide systems into systems
that are what we'll
151
00:10:24,590 --> 00:10:29,420
refer to as time-invariant
or time-varying systems.
152
00:10:29,420 --> 00:10:33,320
And these aren't terms that
we've defined yet, of course,
153
00:10:33,320 --> 00:10:37,160
but we will be defining in
the course very shortly.
154
00:10:37,160 --> 00:10:40,640
And while, in some sense, this
of signals with
167
00:11:27,020 --> 00:11:28,540
an appropriate system.
168
00:11:28,540 --> 00:11:32,630
The particular example that I
want to illustrate relates to
169
00:11:32,630 --> 00:11:35,740
the restoration of
old recordings.
170
00:11:35,740 --> 00:11:38,960
And this is some work that was
done by Professor Thomas
171
00:11:38,960 --> 00:11:43,190
Stockham, who is at the
University of Utah, and work
172
00:11:43,190 --> 00:11:46,800
that he had done a number of
years ago relating to the fact
173
00:11:46,800 --> 00:11:53,380
that in old recordings, for
example in Caruso recordings,
174
00:11:53,380 --> 00:11:57,380
the recording was done through
a mechanical horn, and the
175
00:11:57,380 --> 00:11:59,990
characteristics of the
horn tended to
176
00:11:59,990 --> 00:12:01,860
vary from day to day.
177
00:12:01,860 --> 00:12:04,900
And because of the
characteristics of the horn,
178
00:12:04,900 --> 00:12:08,190
217
00:14:33,020 --> 00:14:37,580
the background noise on the
recording was emphasized.
218
00:14:37,580 --> 00:14:43,310
And so he processed the signal
further in an attempt to
219
00:14:43,310 --> 00:14:45,350
remove that background noise.
220
00:14:45,350 --> 00:14:48,250
And with that particular
processing, the processing was
221
00:14:48,250 --> 00:14:51,060
very highly nonlinear.
222
00:14:51,060 --> 00:14:54,980
A very interesting thing
happened, which was that not
223
00:14:54,980 --> 00:14:59,060
only in that processing was the
background noise removed,
224
00:14:59,060 --> 00:15:02,180
but somewhat surprisingly,
also the
225
00:15:02,180 --> 00:15:03,630
orchestra was removed.
226
00:15:03,630 --> 00:15:08,240
And let me just play that now
as an example of some very
227
00:15:08,240 --> 00:15:11,930
sophisticated processing with
a nonlinear system.
228
00:15:11,930 --> 00:15:17,515
What you'll hear on channel one
is the restoration as we
229
00:15:17,515 --> 00:15:19,320
had just played it.
230
00:15:19,320 --> 00:15:23,610
When I switch to channel two,
it will be after the
231
00:15:23,610 --> 00:15:27,986
processing with an attempt to
remove the orchestra and the
232
00:15:27,986 --> 00:15:30,060
background noise.
233
00:15:30,060 --> 00:15:31,310
Channel one now.
234
00:15:31,310 --> 00:15:38,696
235
00:15:38,696 --> 00:15:41,691
And now the noise and
orchestra removed.
236
00:15:41,691 --> 00:15:44,390
237
00:15:44,390 --> 00:15:46,293
Back to channel one.
238
00:15:46,293 --> 00:15:58,460
239
00:15:58,460 --> 00:16:02,085
And finally, once again, with
the orchestra removed.
240
00:16:02,085 --> 00:16:12,220
241
00:16:12,220 --> 00:16:15,950
So that's an example
of processing of a
242
00:16:15,950 --> 00:16:19,280
continuous-time signal
with a corresponding
243
00:16:19,280 --> 00:16:21,160
continuous-time system.
244
00:16:21,160 --> 00:16:25,800
Now I'd like to illustrate an
example of some processing on
245
00:16:25,800 --> 00:16:27,600
a discrete-time signal.
246
00:16:27,600 --> 00:16:30,730
And I'd like to do that in the
context of the example that I
247
00:16:30,730 --> 00:16:33,260
showed before of a discrete-time
signal, which
248
00:16:33,260 --> 00:16:38,660
was the Dow Jones Industrial
weekly stock market index.
249
00:16:38,660 --> 00:16:44,880
I had shown it before, as I've
shown it here again, over a
250
00:16:44,880 --> 00:16:49,550
period of slightly more than a
year, where this is the number
251
00:16:49,550 --> 00:16:53,630
of weeks and this is
the weekly index.
252
00:16:53,630 --> 00:16:58,640
And to illustrate some of the
processing, what I'd like to
253
00:16:58,640 --> 00:17:03,020
do is show the stock market
index, the weekly index, over
254
00:17:03,020 --> 00:17:07,750
278
00:18:42,325 --> 00:18:47,560
And in particular, what's done
is to smooth it, or average
279
00:18:47,560 --> 00:18:53,820
over some time period, to
emphasize the slow variations
280
00:18:53,820 --> 00:18:56,750
and de-emphasize the
rapid variations.
281
00:18:56,750 --> 00:19:01,070
And that, in fact, is processing
that is done with a
282
00:19:01,070 --> 00:19:02,560
discrete-time system.
283
00:19:02,560 --> 00:19:06,660
So when you hear referred to,
let's say, in stock market
284
00:19:06,660 --> 00:19:11,070
reports, a 51-day moving
average, that, in fact, is
285
00:19:11,070 --> 00:19:15,630
processing the stock market
index with a particular
286
00:19:15,630 --> 00:19:17,460
discrete-time system.
287
00:19:17,460 --> 00:19:21,960
The result of doing that on
this particular example
288
00:19:21,960 --> 00:19:25,090
generates a smooth version
of the curve,
289
00:19:25,090 --> 00:19:27,040
which I overlay here.
290
00:19:27,040 --> 00:19:34,330
And the overlay, then, is really
attempting to track the
291
00:19:34,330 --> 00:19:37,600
smoother variations and
de-emphasize the more rapid
292
00:19:37,600 --> 00:19:39,370
variations.
293
00:19:39,370 --> 00:19:42,930
Let me just slightly offset that
so that the difference
294
00:19:42,930 --> 00:19:44,680
stands out a little more.
295
00:19:44,680 --> 00:19:51,230
And so here you see what is
the original weekly index.
296
00:19:51,230 --> 00:19:56,280
And this is the result of
processing that sequence with
297
00:19:56,280 --> 00:20:00,990
an appropriate system
to apply smoothing.
298
00:20:00,990 --> 00:20:04,110
And in fact, what it is
is a moving average.
299
00:20:04,110 --> 00:20:08,330
And so here again, you can see,
in the smoother curve,
300
00:20:08,330 --> 00:20:12,800
this general downward trend
up until this time period,
301
00:20:12,800 --> 00:20:17,620
followed by, eventually,
a recovery.
302
00:20:17,620 --> 00:20:23,560
Well, we've seen an example with
a continuous-time signal,
303
00:20:23,560 --> 00:20:26,070
the Caruso recording, an example
of the discrete-time
304
00:20:26,070 --> 00:20:28,900
signal, this stock
market index.
305
00:20:28,900 --> 00:20:33,810
And what I'd also like to show
is a third example, which is
306
00:20:33,810 --> 00:20:38,320
the result of some processing on
an image, in particular the
307
00:20:38,320 --> 00:20:40,720
image that we talked about
before, which was the aerial
308
00:20:40,720 --> 00:20:45,000
photograph that had the problem
of some cloud cover.
309
00:20:45,000 --> 00:20:50,570
So once again, what we see here
is the original aerial
310
00:20:50,570 --> 00:20:53,770
photograph with the
cloud cover.
311
00:20:53,770 --> 00:20:58,260
And some processing was applied
to this using a system
312
00:20:58,260 --> 00:21:02,520
which, in fact, was both
nonlinear and quote
313
325
00:21:52,790 --> 00:21:57,080
continuous, as they are in a
spatial, aerial photograph, in
326
00:21:57,080 --> 00:22:01,180
fact, for the processing, that
picture was first converted to
327
00:22:01,180 --> 00:22:05,050
a sequence through a process
called sampling, which we'll
328
00:22:05,050 --> 00:22:06,620
be talking about later.
329
00:22:06,620 --> 00:22:09,180
And then the processing,
in fact, was done
330
00:22:09,180 --> 00:22:10,430
on a digital computer.
331
00:22:10,430 --> 00:22:12,890
332
00:22:12,890 --> 00:22:19,340
Well, these then are some
examples of the use of some
333
00:22:19,340 --> 00:22:25,490
systems to process some signals,
both in continuous
334
00:22:25,490 --> 00:22:30,290
time and discrete time, for
one-dimensional signals and
335
00:22:30,290 --> 00:22:34,190
for multi-dimensional signals.
336
00:22:34,190 --> 00:22:38,580
And as I've referred to systems,
we've thought of them
337
349
00:23:21,340 --> 00:23:27,280
Feedback, as you may or may not
know, comes into play in a
350
00:23:27,280 --> 00:23:31,250
variety of situations, for
example, in amplifier design,
351
00:23:31,250 --> 00:23:36,770
as we'll talk about, feedback
plays an important role.
352
00:23:36,770 --> 00:23:42,310
In a situation where you have
a basically unstable system,
353
00:23:42,310 --> 00:23:45,320
feedback is often used to
stabilize the system.
354
00:23:45,320 --> 00:23:49,520
And feedback interconnections
of systems in that sense are
355
00:23:49,520 --> 00:23:53,020
very often used in high
performance aircraft, which
356
00:23:53,020 --> 00:23:56,830
are inherently unstable, and
are stabilized through this
357
00:23:56,830 --> 00:23:59,770
kind of interconnection.
358
00:23:59,770 --> 00:24:03,760
Just to give you a little sense
of this without going
359
00:24:03,760 --> 00:24:08,970
into any of the details, what
I'd like to show you is an
360
00:24:08,970 --> 00:24:15,490
excerpt from a lecture that
course to be educational
457
00:30:18,910 --> 00:30:19,720
and interesting.
458
00:30:19,720 --> 00:30:22,250
And I certainly hope that
that will be the case.
459
00:30:22,250 --> 00:30:23,500
Thank you.
460
00:30:23,500 --> 00:30:25,045